Introduction
Episode 1: From text to vectors
Figure 1
Embedding of a cat - We have described it along
two dimensions: furriness and number of legs
Figure 2
Embeddings of a cat and a dog and a
caterpillar
Figure 3
Figure 4
Embeddings of a cat and a dog and a caterpillar
- We can describe these animals in many dimensions!
Figure 5
Schematic representations of the different
prediction tasks that CBOW and Skip-gram try to solve
Episode 2: BERT and TransformersTransformersBERTBERT ArchitectureBERT as a Language ModelBERT for Text ClassificationUnderstanding BERT ArchitectureBERT for Token Classification———- END HERE ??? ———-
Figure 1
Transformer Architecture
Figure 2
BERT Architecture
Figure 3
BERT Language Modeling
Figure 4
BERT as an Emotion Classifier
Figure 5
BERT as an Emotion Classifier
Figure 6
The Encoder-Decoder Attention Mechanism
Figure 7
The Encoder Self-Attention Mechanism
Figure 8
BERT as an NER Classifier